Nothing to paste? — a real-shaped scrap heap for a churn classifier: partial frontmatter with no licence, two template placeholders still in it, an AUC with no evaluation set, and a dataset named but never sourced. Or see a finished card with no run and no charge: or .
Paste the material
An existing card, a training config, an eval table, a dataset note — in any order and any state. The quick linter reads the frontmatter, checks the licence id, maps your headings onto the Hub template and counts placeholders and unbacked metric claims, free in your browser. No sign-in needed for that.
Get the card
Hub-ready YAML frontmatter, model details, intended and out-of-scope uses drawn from what the data actually covers, training and evaluation with every metric tied to its dataset and split, and limitations that follow from the material rather than from boilerplate. Every backticked value is checked back against what you pasted, and anything not found is marked unverified instead of shown as fact.
Close the gaps
Gaps ranked Blocker, Should fix or Nice to have, each naming the evidence and exactly what to go and fetch — plus the claims a reader could draw that your material does not support. Take the README.md, close the blockers, paste the richer material and run it again.